Identifying Latent Cognitive Constructs in a Comprehensive Model of Aviation Outcomes: The Role of the Dynamic Mental Model for Pilots

Background: Older private pilots are at greater risk for general aviation accidents when compared with younger pilots. Predicting which older pilots are vulnerable to accidents has proven elusive, despite the inclusion of domain-independent cognitive factors, such as working memory and processing speed, in outcome models. Thus, it has been suggested that domain-dependent cognitive factors should feature in models of pilot performance. The mental model for pilots was highlighted as a domain-dependent variable particularly relevant to explanations of pilot performance (Hardy & Parasuraman, 1997). The present study specified a mental model for pilots and examined its capacity to predict aviation outcomes. Methods: A flight simulation program of research collected performance data for 108 pilots who “flew” a Cessna 172. The findings from the flight simulation research produced a robust set of data structures utilized in the modeling phases of this work. A dynamic mental model for pilots (DMM) was developed with guidance from a framework first proposed by Herdman and LeFevre (2004). Structural equation modeling was used to validate the DMM as a central element in a cognition-oriented model of pilot performance for aviation (COMP-A). The COMP-A, based on work by Hardy and Parasuraman (1997), was used as a platform for quantifying the effects of pilot attributes and cognition on aviation outcomes. Findings: Confirmatory factor analysis demonstrated the DMM convergence as a formative-reflective construct composed of domain-relevant components. The COMP-A included an expansive selection of predictor variables and outcomes of relevance to aviation: critical incidents and runway incursion management. The DMM scores provided the best explanation of critical incidents. The DMM also mediated the majority of effects from upstream variables, such as pilot age, on aviation outcomes and was the only variable to predict risk of a critical incident with reasonable sensitivity and specificity. Implications: Findings challenge current strategies for assessing the aging aviator. First, visual-spatial attention, and not working memory or processing speed, was the domain-independent variable with the largest explanatory power in the COMP-A. Second, domain-dependent, not domain-independent, constructs were the most efficacious predictors of risk.